Cone-beam CT (CBCT) is finding increased use in image-guided procedures, including orthopaedic surgeries such as spine fusion and total hip arthroplasty. Since intraoperative imaging is particularly likely to include surgical devices (e.g. tools, implants, or prostheses) within the tomographic field-of-view and these components have known composition, size, and shape, there is a unique opportunity to integrate such information in a re- construction approach. The investigators have developed a novel model-based approach called known- component reconstruction (KCR) that leverages known attenuation distributions, modeling an object comprised of known components (with unknown pose), as well as an unknown background anatomy. This is a new paradigm for incorporating prior object knowledge into a reconstruction framework where the algorithm jointly estimates both the background attenuation and the registers the known components. The technique is particularly well-suited to missing data and low signal-to-noise, as is common in interventional imaging due to metallic devices. Traditional reconstruction approaches are prone to severe metal streak artifacts (especially at low doses) with the poorest image quality in locations proximal to the device, which is often precisely the area of interest with the greatest image quality demands (e.g. visualization of nearby critical structures or interfaces of implants). Preliminary studies demonstrate that KCR is able to essentially eliminate artifacts associated with metal and allows for visualization of the object right up to the boundary of the tool or implant. We hypothesize tha an integrated system based on a generalized KCR framework with a library of known device components can provide artifact-free reconstructions in proximity to surgical implants, facilitatin high- precision device placement and dose reduction protocols in interventional CBCT. The following Specific Aims are proposed: 1.) Build a generalized analytic framework for KCR. Studies include development of a complete physics model for interventional CBCT, leveraging KCR's unique integration of component know- ledge, and adopting a deformable transformation model to allow for a broad class of inexactly known components (e.g., fixation rods in spine fusions that are deformed during a procedure to enforce a specific spine curvature). 2.) Create an integrated system for KCR. The development includes methods for generation of high- fidelity parameterized component models from CAD files or physical devices, computationally efficient algorithms and hardware, and tools for assessment of geometric accuracy in device placement from the component registration computed jointly in KCR. 3.) Evaluate KCR in pre-clinical experiments and simulated procedures. Work includes a systematic series of experiments using phantoms and cadavers with multiple components, deformable constructs, and conditions that stress the limits of noise, dose, object size, and implant size. Outcome measures will include quantitative imaging performance metrics, physician scoring, and registration error analysis, as well as the relation of these metrics to minimum-dose acquisition protocols.

Public Health Relevance

3D-capable C-arms are used for an increasing number of image-guided orthopaedic procedures that involve the introduction of metallic surgical devices into the field-of-view, including spine fusions and total hip arthroplasty. Achieving high quality images in the presence of such components is severely challenged in conventional computed tomography (CT), particularly in close proximity to the metal - which is often the area of critical interest to the surgeon in visualizing anatomy immediately around the implant. The challenge to image quality becomes even greater in light of the important push to minimize medical-imaging-related radiation dose for both patients and surgical staff. The research proposed herein involves a novel system for CT image reconstruction that directly incorporates knowledge about the specific surgical devices that are present in the patient, precisely analyzes their location within the body, and simultaneously computes a model- based image reconstruction that is nearly artifact-free and allows high-quality visualization of the implant and surrounding tissue right up to the boundary of the metal implant, providing a major advance in image quality, facilitating the drive to dose reduction, and giving quantitative feedback on the geometric accuracy of device placement.

National Institute of Health (NIH)
National Institute of Biomedical Imaging and Bioengineering (NIBIB)
Exploratory/Developmental Grants (R21)
Project #
Application #
Study Section
Biomedical Imaging Technology Study Section (BMIT)
Program Officer
Krosnick, Steven
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Johns Hopkins University
Biomedical Engineering
Schools of Medicine
United States
Zip Code
Tilley 2nd, Steven; Siewerdsen, Jeffrey H; Stayman, J Webster (2014) Iterative CT Reconstruction using Models of Source and Detector Blur and Correlated Noise. Conf Proc Int Conf Image Form Xray Comput Tomogr 2014:363-367
Stayman, J Webster; Zbijewski, Wojciech; Tilley 2nd, Steven et al. (2014) Generalized Least-Squares CT Reconstruction with Detector Blur and Correlated Noise Models. Proc SPIE Int Soc Opt Eng 9033:903335
Stayman, J Webster; Tilley 2nd, Steven; Siewerdsen, Jeffrey H (2014) Integration of Component Knowledge in Penalized-Likelihood Reconstruction with Morphological and Spectral Uncertainties. Conf Proc Int Conf Image Form Xray Comput Tomogr 2014:111-115